| Abstrakt: |
One of the most common diseases in recent years has been lung cancer. In the US, around 200,000 new instances are reported annually, based on studies in this area. Malignant tumors are created when lung cells proliferate and expand out of control. Convolutional Neural Networks (CNN), in particular, are deep learning algorithms that have emerged as the best method for autonomously diagnosing illness in recent times. In the literature, some methods are reviewed, but they do not provide efficient detection. A few techniques are evaluated in the literature, however, they don't offer effective detection. To classify lung cancer, the Optimal DenseNet is therefore built in this study. DenseNet and Modified Golf Optimization (MGO) were used in the creation of this structure. The MGO selects the best hyperparameters for fully connected layers (number of hidden units, number of fully connected layers) and convolution layers (size and number of filters). Initially, lung cancer images were gathered and taken into consideration for the data analysis procedure. The next step is feature extraction, which uses wavelet, texture, and histogram characteristics to extract the important information from the enhanced images. The categorization stage receives the features. Using the image, lung cancer is classified as either malignant or non-cancerous at the classification step. MATLAB is used to implement the suggested approach, and metrics are calculated for validation. The method's performance is justified by drawing comparisons with conventional approaches. |
Full text from SpringerLink